Variational Approximations for Generalized Linear Latent Variable Models

Generalized linear latent variable models (GLLVMs) are a powerful class of models for understanding the relationships among multiple, correlated responses. Estimation, however, presents a major challenge, as the marginal likelihood does not possess a closed form for nonnormal responses. We propose a variational approximation (VA) method for estimating GLLVMs. For the common cases of binary, ordinal, and overdispersed count data, we derive fully closed-form approximations to the marginal log-likelihood function in each case. Compared to other methods such as the expectation-maximization algorithm, estimation using VA is fast and straightforward to implement. Predictions of the latent variables and associated uncertainty estimates are also obtained as part of the estimation process. Simulations show that VA estimation performs similar to or better than some currently available methods, both at predicting the latent variables and estimating their corresponding coefficients. They also show that VA estimation offers dramatic reductions in computation time particularly if the number of correlated responses is large relative to the number of observational units. We apply the variational approach to two datasets, estimating GLLVMs to understanding the patterns of variation in youth gratitude and for constructing ordination plots in bird abundance data. R code for performing VA estimation of GLLVMs is available online. Supplementary materials for this article are available online.     

Superconvergent Recovery of the Gradient from Piecewise Linear Finite-element Approximations

We propose and justify the use of a simple scheme which recoversgradients from the piecewise linear finite-element approximationon triangular elements to the solution of a second-order ellipticproblem. The recovered gradient is a superconvergent estimateof the true gradient at the midpoints of element edges. A relatedscheme recovers the gradient at the element centroids.     

分段线性插值收敛速度的一种估计

给出了分段线性插值收敛速度的一种估计.     

Estimates for Piecewise Quadric Approximations of Implicitly Defined Manifolds

Several algorithms have been given for obtaining piecewise quadric approximations of implicitly defined manifolds. This paper is concerned with error estimates for such algorithms.     

Models and Approximations for Call Center Design

A call center is a facility for delivering telephone service, both incoming and outgoing. This paper addresses optimal staffing of call centers, modeled as M/G/n queues whose offered traffic consists of multiple customer streams, each with an individual priority, arrival rate, service distribution and grade of service (GoS) stated in terms of equilibrium tail waiting time probabilities or mean waiting times. The paper proposes a methodology for deriving the approximate minimal number of servers that suffices to guarantee the prescribed GoS of all customer streams. The methodology is based on an analytic approximation, called the Scaling-Erlang (SE) approximation, which maps the M/G/n queue to an approximating, suitably scaled M/G/1 queue, for which waiting time statistics are available via the Pollaczek-Khintchine formula in terms of Laplace transforms. The SE approximation is then generalized to M/G/n queues with multiple types of customers and non-preemptive priorities, yielding the Priority Scaling-Erlang (PSE) approximation. A simple goal-seeking search, utilizing SE/PSE approximations, is presented for the optimal staffing level, subject to GoS constraints. The efficacy of the methodology is demonstrated by comparing the number of servers estimated via the PSE approximation to their counterparts obtained by simulation. A number of case studies confirm that the SE/PSE approximations yield optimal staffing results in excellent agreement with simulation, but at a fraction of simulation time and space.     

Linear and Discontinuous Approximations for Optimal Control Problems

Abstract

An optimal control problem for 2D and 3D elliptic equations is investigated with pointwise control constraints. This paper is concerned with the discretization of the control by piecewise linear but discontinuous functions. The state and the adjoint state are discretized by linear finite elements. The paper is focused on similarities and differences to piecewise constant and piecewise linear (continuous) approximation of the controls. Approximation of order h in the L ^∞-norm is proved in the main result.     

On Accurate Linear Approximations for Chance-Constrained Programming

This paper introduces the CHAPS (chance-constrained programming system) algorithm, which uses linearization techniques but gives more accurate solutions than earlier, similar methods.     

Convergence Rate for LS Estimator in Simple Linear EV Regression Models

In this paper, we consider the following simple linear errors-in-variables regression model ${\eta_i=\theta+\beta x_i+\varepsilon_i, \xi_i=x_i+\delta_i, 1\leq i\leq n }$ . The exponential convergence rate for the least squares estimators of the unknown parameters θ, β in this model are obtained.     

Data Dependent Triangulations for Piecewise Linear Interpolation

Given a set of data points in R² and corresponding data values,it is clear that the quality of a piecewise linear interpolationover triangles depends on the specific triangulation of thedata points. While conventional triangulation methods dependonly on the distribution of the data points in R² in this paperwe suggest that the triangulation should depend on the datavalues as well. Several data dependent criteria for definingthe triangulation are discussed and efficient algorithms forcomputing these triangulations are presented. It is shown fora variety of test cases that data dependent triangulations canimprove significantly the quality of approximation and thatlong and thin triangles, which are traditionally avoided, aresometimes very suitable.     

The Optimal Approximations for Solving Linear Ill-Posed Problems

New projection discrete schemes for ill-posed problems are constructed. We show that for equations with self-adjoint operators the use of self-adjoint projection schemes is not optimal in the sense of the amount of discrete information.     

具有分片线性NCP函数的QP-Free可行域方法

In this paper, a QP-free feasible method with piecewise NCP functions is proposed for nonlinear inequality constrained optimization problems. The new NCP functions are piecewise linear-rational, regular pseudo-smooth and have nice properties. This method is based on the solutions of linear systems of equation reformulation of KKT optimality conditions, by using the piecewise NCP functions. This method is implementable and globally convergent without assuming the strict complementarity condition, the isolatedness of accumulation points. Furthermore, the gradients of active constraints are not requested to be linearly independent. The submatrix which may be obtained by quasi-Newton methods, is not requested to be uniformly positive definite. Preliminary numerical results indicate that this new QP-free method is quite promising.     

Approximations for Time-Dependent Distributions in Markovian Fluid Models

In this paper we analyse Markov-modulated fluid processes over finite time intervals. We study the joint distribution of the level at time \(\theta < \infty \) and of the maximum level over [0, θ], as well as the joint distribution of the level at time θ and the minimum level over [0, θ]. We approximate θ by a random variable T with Erlang distribution and so use an approach different from the usual Laplace transform to compute the distributions. We present probabilistic interpretation of the equations and provide a numerical illustration.     

First Passage Time for Brownian Motion and Piecewise Linear Boundaries

We propose a new approach to calculating the first passage time densities for Brownian motion crossing piecewise linear boundaries which can be discontinuous. Using this approach we obtain explicit formulas for the first passage densities and show that they are continuously differentiable except at the break points of the boundaries. Furthermore, these formulas can be used to approximate the first passage time distributions for general nonlinear boundaries. The numerical computation can be easily done by using the Monte Carlo integration, which is straightforward to implement. Some numerical examples are presented for illustration. This approach can be further extended to compute two-sided boundary crossing distributions.     

Asymptotic Directions for Best Approximations of n-Dimensional Linear Forms

Some results about asymptotic directions for best approximations of linear forms similar to known results on best simultaneous approximations are obtained.     

����������ĸ����¼��°��������ģ��

??Recurrent event data usually occur in long-term studies which concern
recurrence rates of the disease. In studies of medical sciences, patients who have infected
with the disease, like cancer, were conventionally regarded as impossible to be cured. However,
with the development of medical sciences, recently those patients were found to be possibly
recovered from the disease. The recurrence rate of the events, which is of primary interest,
may be affected by the cure rate that may exist. Therefore, we proposed semiparametric
statistical analysis for recurrent event data with subjects possibly being cured. In our
approach, we present a proportional rate model for recurrence rate with the cure rate adjusted
through a Logistic regression model, and develop some estimating equations for estimation of
the regression parameters, with their large sample properties, including consistency and
asymptotic normality established. Numerical studies under different settings were conducted
for assessing the proposed methodology and the results suggest that they work well for
practical situations. The approach is applied to a bladder cancer dataset which motivated our
study.     

Exact Penalty and Optimality Condition for Nonseparable Continuous Piecewise Linear Programming

Utilizing compact representations for continuous piecewise linear functions, this paper discusses some theoretical properties for nonseparable continuous piecewise linear programming. The existence of exact penalty for continuous piecewise linear programming is proved, which allows us to concentrate on unconstrained problems. For unconstrained problems, we give a sufficient and necessary local optimality condition, which is based on a model with universal representation capability and hence applicable to arbitrary continuous piecewise linear programming. From the gained optimality condition, an algorithm is proposed and evaluated by numerical experiments, where the theoretical properties are illustrated as well.